Chapter 7 of 17
Gold flows, positioning and point-in-time research
Reconcile ETF metal holdings, fund value, vault stocks and positioning without double counting or using information before publication.
In this chapter
Separate quantities from values #
An increase in gold-fund AUM can reflect a higher metal price with unchanged holdings. A metal-tonnage change is a different measure from a dollar-flow estimate, and a unit creation can have specific cash or in-kind mechanics. Read the provider’s methodology rather than deriving every flow from a market-value change.
The World Gold Council’s dataset aggregates specified physically backed products and reports its definitions. Treat the coverage and reporting date as part of the observation. A revised figure is not a number that a historical researcher could necessarily have known at the earlier date.
Do not add overlapping stocks #
A vault series is a location-based stock measure; an ETF holding is an ownership/product measure. Where product holdings sit inside the reported vault perimeter, adding them to vault stocks counts the same metal twice. That does not make either dataset false; it means their dimensions differ.
Build an overlap map with location, owner/product, custody chain, unit and reporting cutoff. If overlap cannot be established from public data, report the uncertainty instead of presenting a precise “free float” estimate.
Positioning is an imperfect exposure map #
CFTC disaggregated categories classify traders, not every trade’s motive. A producer-category position is not automatically proof of informed directional selling, and a managed-money position is not a complete map of all global speculative exposure. The futures-only and combined report types are distinct datasets.
Keep the report’s observation date and public release timestamp. The usual Tuesday/Friday lag is not permission to assume every week follows the same calendar; consult the actual release schedule and revisions. A Friday publication cannot be used as a Tuesday trading signal in a point-in-time test.
Normalise the numerator and the denominator #
Managed-money net length is longs minus shorts in the specified report. A net/open-interest ratio needs matching report coverage, units and date. A rise in the ratio may come from a falling denominator rather than new longs. Inspect gross longs, gross shorts and spreading separately.
Percentile ranks depend on the sample and any changes in classifications or market structure. Register the window and the rule for missing observations before measuring a signal. A historically extreme position can remain extreme; it is not a preassigned turning date.
Ask whether the signal adds information #
A flow variable observed after price has moved may partly reflect that movement. Compare lagged flow, contemporaneous return and the information available at the proposed decision time. Use an appropriately frozen model and evaluation period rather than narrating selected examples.
Report whether a candidate improves prediction or risk control after costs and competing controls, not only whether the sign matches a story. No public dataset supplies a complete institutional gold book, and confident language cannot close that coverage gap.
Hypothetical worked example
Two fictional reconciliation traps
A fund has 100,000 ounces worth 200 million dollars at 2,000 dollars per ounce. Price rises 10% with no quantity change: value becomes 220 million, but the 20 million increase is not a new 20 million subscription.
In another dataset, net reported length stays at 20,000 contracts while matching total open interest falls from 100,000 to 80,000. Net/open interest rises from 20% to 25% without net buying. Both the numerator and the denominator belong in the research note.
Common mistakes
- Calling AUM appreciation a subscription.
- Adding custody stocks and overlapping product holdings.
- Trading a COT observation before its public release.
A fund’s AUM rises while its metal quantity is unchanged. Which conclusion is justified?
Make this chapter useful
Can you defend the reasoning?
Before moving on, answer the self-check in your own words and write why each alternative fails. Use the worked example to test that explanation. Then apply it in the module research task; keep the assumptions and the evidence that could change your answer.
Desk assignment
What to submit
Prepare a point-in-time source ledger and overlap matrix. Write one hypothesis that the available public data can test and one it cannot.
Use the research ledger for dated evidence, or the module workbooks for calculations. Separate observations, assumptions and interpretation.
Sources and further reading
- World Gold Council: physical gold ETF holdings and flows ↗
- LBMA: London vault data and reporting scope ↗
- CFTC: disaggregated positioning classifications and limitations ↗
- CFTC: report release schedule ↗
- World Gold Council: gold market size and structure ↗
Further-reading references from the curriculum source list. Contract sizes, expiry calendars, settlement references, margin parameters and tax treatment can change; verify the current operative documents for the actual product. Worked figures are illustrative, not current quotes.
Progress is stored only on this browser.
Back to Gold: market structure and portfolio risk · Learning library · Calculators
General financial education. No securities recommendation, assured outcome or personal advice. Read the risk disclosures.